AUTOMATIC SOLDER PASTE PRINTER POSTIONAL FEEDBACK

CONTROL

Paul Haugen.
CyberOptics Corp
Minneapolis, MN

Rita Mohanty Ph.D

Speedline Technologies
Franklin, MA

ABSTRACT
Stencil printing is a critical first step in surface mount
assembly. It is often cited that the solder paste printing
operation causes about 50%-80% of the defects found in the
assembly of PCBs. Printing is widely recognized as a
complex process whose optimal performance depends on
the adjustment of a substantial number of parameters. It is
not uncommon to hear that stencil printing is more of an
“art” than “science”. In fact, the process is so complex that
sub-optimal print parameters usually end up being used. In
addition, stencil printing produce relatively noisy data,
which makes the print process extremely difficult to control.
Ideally, minimizing the variance of the deposited location
and volumes will improve the quality of the process and
produce more reliable solder joints.
In general, there are two critical aspects to a printing
process. The goal of solder paste printing is to put down the
right “volume” of paste on the right “spot”. In another
words, we not only have to monitor the amount of paste
volume we also need to monitor X, Y and Ө registration of
the board. This issue is compounded when dealing with
miniature components such as 0201, 01005, 0.4mm and
0.3mm CSP and lead free paste. Lead-free paste is known to
have less spread, or wet-ability, and adds to the challenge.
In this paper we present the results of tests that show
registration control, using a closed loop control scheme, is
viable. Using registration feedback from a solder paste
inspection system to the solder paste screen printer, the
solder paste can be repeatedly put in the right “spot” board
after board. We will also discuss best practices regarding
how to set up the closed loop between the solder paste
inspection system and the screen printer. To develop a
robust feedback mechanism, the control process must work
over a wide range of board layouts and configurations. To
this end, there are best practices to follow when setting up
the control loop that optimize the feedback process. To
define these best practices, we will discuss the sensitivity
the feedback loop parameters such as the board layout, paste
location, and the solder paste inspection setup.
Key words: Closed loop process control, solder paste
inspection, print registration, miniature components, process
yield, feedback loop..
INTRODUCTION
Inspections systems for SMT lines have been mainly used
for defect detection and process characterization. Although
these are important functions for an inspection machines, it
is usually better to control a process thus improving the
yield and throughput of an SMT lines than it is to detect
defects after the defect has been generated. Closed-loop
process controls are starting to appear in SMT
manufacturing lines, as board assemblers are demanding
higher first pass yields with higher quality.
Closed loop process control can be defined as a system that
continually monitors and adjusts a process to maintain a
particular target value of an output or outputs. For a closed-
loop process control system to function we must identify the
output or outputs factors and what input factors influence
the variation of those outputs. For solder paste printing you
may, for example, consider paste deposit height, volume,
shape, etc. as the outputs for a solder paste printing process
and print process parameters, paste type, tooling etc. as the
inputs. Once these factors are identified, you must then
quantify these factors through the use of formal statistical
tools such as Design of Experiments (DOE) to fully realize
the benefits.
Presently, closed loop process control is primarily seen in
the component placement and reflow processes. As the
boards are getting denser and components are getting
smaller, most assemblers are looking for ways to prevent
defects before they impact yields. One area where process
control can be implemented with significant impact to the
entire assembly process is the solder paste printing process.
We are starting to see genuine interest in a defect prevention
capability rather than defect detection capability. It is well
understood that the volume of paste deposit is a direct
function of the stencil thickness, but is influenced by other
process parameters outside the control of the printer making
volume control a challenge. However, positional accuracy
control, X, Y and θ can be directly controlled by the printer.
Controlling positional accuracy is the first step in
controlling the overall print process. It is a first step in
realizing the full potential of solder paste inspection
systems. In the following sections, we discuss a solder
paste printing feedback method that improves the alignment
capability of the printing process
Copper Pad PCB

POSITIONAL ACCURACY CONTROL

Positional or registration control is possibly the most
straight forward closed-loop control system that can be
applied to a solder paste printer. Although most printers
have automated systems that perform the alignment of the
stencil to the PCB, it is not uncommon to see solder paste
deposits end up at locations that are not ideal. Figure 1a, b
and c demonstrates such scenarios. It is clear for the figures
below, condition ‘c’, even though delivering correct amount
of paste, will be detrimental to a board assembly process.
Solder paste alignment errors can be produced either by Solder
Paste
board-to-board variations, stencil stretch, inaccuracies in the
alignment system or from other sources.

Figure 1c. Clearly mis-aligned solder paste that would

PCB most likely cause assembly errors.
Copper Pad

In general, the position of the deposits relative to the pads

can be measured within the printer. However, since the
inspection is typically carried out while the board is still
inside the printer, this operation can considerably slow
down the primary function of the printer, resulting in lower
throughput. In addition, if there is a mis-alignment or
calibration issue with the vision system within the printer,
the error is likely not detected since the same vision system
is used to align the stencil as it is used to measure the solder
paste position. Both these issues are addressed by using an
Solder external solder paste inspection system. Subsequently, the
Paste
inspection result can be fed back to the printer to make
required offset adjustment to keep the process in control.

Figure 1a. Perfectly aligned solder paste To improve the performance of the positional accuracy we
tested a closed loop system using a CyberOptics SE300 SPI
system to measure X, Y and θ print offset of solder paste on
Copper Pad PCB a PCB. This information was then passed on back to the
printer which made appropriate corrections to the print
offset as necessary. Figure 2 shows the closed loop concept
and figure 3 details the actual data flow through the
feedback loop.

Solder
Paste

Figure 1b. Partially mis-aligned solder paste that may be

acceptable in many SMT line

Figure 2 Closed loop print control concept


Figure3. Details of the closed loop feed back loop used to
correct and maintain solder paste alignment.
MEASUREMENT OF SOLDER PASTE OFFSET
To correctly apply a solder paste alignment feedback loop,
the positional error of the print must first be defined. The
goal of a positional feed back loop is to correct for any
misplacement of solder paste. In any classical feed back
mechanism, the difference in the measured control value
and the goal for that control value – the error signal- is what
is typically used as feedback. To that end, we will need a
common definition of the “error signal” that is shared
between the solder paste inspection system and the printer.
The first step in specifying a common interface between SPI
and the printer is defining a common positional frame of
reference between the two systems. The most obvious frame
of reference that can be shared between the SPI system and
the printer is the position frame defined by the fiducials
found on the PCB. Fiducials are ubiquitous on all PCBs and
are required feature on the PCB’s copper planes to provide a
common frame of reference used by all automated machines
in a circuit board assembly line. Using the fiducial frame of
reference, a common set of axes (X and Y) can be defined to
form a basis for communicating the mis-alignment of the
solder paste.
The second step is measuring the “error signal”. Figure 4a
shows a CAD defined aperture aligned with a copper pad.
Since the fiducials are usually located on the copper layer,
the alignment of the CAD aperture to the pads is usually
pretty good. Figure 4b shows a solder deposit that has been
perfectly placed on the pad. Both the center of the CAD
aperture and the center of the solder paste deposit can be
calculate using a weighted center-of-gravity algorithm that
determines the center coordinates of either the CAD
aperture or solder paste deposit based on the feature’s shape.
Figure 5a shows a registration of a solder deposit to the
CAD aperture position. Figure 5b shows a similar deposit
that is offset from the CAD aperture. In this case, the “error
signal” would be the resulting difference in the position of
the CAD center and the solder paste center given by X
offset and Y offset in the figure.
However, the position of single deposit does not adequately
reflect the actual alignment of the printer. The actual
position of a single solder paste deposit is affected by the
shape of the deposit, local stretch in the stencil, overprint,
stray solder balls, and slump etc. Because of these affects, it
is not possible to adequately determine the registration of
printer from a single deposit. Also, for single deposits with
rotationally symmetric shape, the notion of rotational (θ)
position error is not defined. Even non-symmetric deposit
shapes such as rectangles, small shape changes such as
overprint and slump will cause more apparent rotation than
reality causing noise in the feedback loop
Figure 4a Figure 4b
Figure 5a Figure 5b
To provide reliable and stable feedback to the printer, the
combined position of a large number of pads is required.
Printers typically have three degrees of freedom for
alignment - X, Y and rotation, θ. Figure 6shows an array of
deposits that show good alignment to the CAD aperture
positions. Figure 6b shows an example of position offset
(both in x and y direction ), Figure 6c shows and example
rotational offset and figure 6d shows both positional and
rotational offset. To effectively align a printer, all three
positional variables must be determined simultaneously and
be fed back to printer as parts of the “error signal”.
X, Y and θ can be determined for the whole PCB by
comparing the center positions (Xs and Ys) of the solder
paste deposits with the center positions of the CAD
apertures (Xc and Yc). Using a least squares optimization
technique, a X, Y and θ offset can be determined that
defines a best fit alignment between the solder paste features
and the CAD features. In terms of the frame of reference
defined by the fiducials, the best fit rotation of the CAD
frame of reference to describe the solder paste position can
be defined with the following equation:
  X s  cos( )  sin( )  X c   X off 
 Y    sin( ) cos( )   Y    Y  (1)
 s    c   off 
where Xoff, Yoff and θ are the data that is passed to the
printer by the solder paste inspection system for each board.
Finding a best fit alignment based on a large population of
solder paste deposits allows the feedback error signal to
directly reflect the capabilities of the printer (i.e. adjusting
for X, Y and θ over the whole stencil). In addition, using
many solder paste deposits reduces the noise in the error
signal returned to the printer.
In addition to the Xoff, Yoff and θ, the center of rotation must
be mutually defined between the solder paste inspection
equipment and the printer. Since rotation is applied before
X and Y offset in the equation above, the values of Xoff and
Yoff are a function of the center of rotation Xcenter and Ycenter.
For our implementation of positional feedback, we chose to
define the center of rotation as the average center of the
solder paste deposits given by
1 n n 1 n n
X center   Xs
n 1
and Ycenter   Ys
n 1
(2)
where Xcenter and Ycenter define the average center of all the
solder paste deposits. Another possible center of rotation
could be the origin of the board or fiducial frame. However,
choosing center of rotation other than Xcenter and Ycenter will
cause the resulting Xoff and Yoff to be large to compensate
any rotation.
EXPERIMENTAL PROCEDURE
To evaluate the performance of the closed loop control,
three tests were conducted. First a gage test was conducted
to determine the capability of the SPI system. Second, a
proof of concept test was run to demonstrate feasibility of
the positional feedback control and to determine the affect
of feedback parameter. Finally, using a production
representative board, we tested the alignment of the printer
with and without feedback control to determine if there is
any improvement in the printer’s process capability when
using feedback control. The following sections detail the
tests and the results.
Gage Testing and Results
Using the board shown in figure 8, a gage study was
conducted to determine the repeatability of the SPI machine.
For the gage study one board was inspected 15 times and the
result is presented in figure 7. The range for the 15 boards
was measured to be less than 5µm for X-Y offset and 0.0020
for Ө. This was considered to be acceptable. Based on this
results, the X, Y offset specification was set to be ±
0.05mm.

Figure 7a. Gage repeatability tests for X and Y offset

measurements

Figure 6a. Perfect alignment Figure 6b. X and Y offset

Figure 7b. Gage repeatability tests for Ө offset

Figure 6c. Rotation offset Figure 6d. X, Y and θ offset
In summary, to implement the feedback loop, the offset,
(Xoff and Yoff ), the center of rotation (Xcenter and Ycenter ),
and rotation θ are passed from the solder paste measurement
system to the printer.
measurements.
Proof of Concept Test Procedure
The objective of this test was to evaluate the performance of
the closed loop controller interface and acceptable range of
correction factor application. The correction factor is
defined as the percentage of SPI measured offset error value
applied to the next board in the printer as a correction. The
correction factor can be considered the proportional gain of
the feedback loop. For example, a 50% correction factor
multiplies the SPI measured alignment offset error by 0.5 to
determine the offset correction that is applied to the next
print. As in any feedback system, optimizing the feedback
loop gain is a compromise between the speed and stability
of the feedback system. For the purposes of this test,
correction factors of 50% and 25% were chosen.

This test was conducted using a standard Speedline test

board which is a 254mm x 203mm x 1.575mm, four layer
FR-4 board with ENIG surface finish. The layout of the Figure 9a. X and Y offset feedback correction with a 50%
board is shown in figure 8. The test board incorporated a correction factor applied to the feedback loop
wide range of commercially available components and
packages such as QFPs, BGAs, PLCCs, QFNs, 0402s, and
0201s. The print test was run on 40 boards to get statistical
confidence in the results.

Figure 9b. Ө offset feedback correction with a 50%

correction factor applied to the feedback loop

Figure 8. Layout of test board used to measure the

capability of the SPI system

Proof of Concept Test Results

Figure 9 and 10 shows the results from the proof of
concepts tests. Results from the rear to front (R2F) squeegee
direction are shown here with two different correction
factors. There results for the front to rear squeegee direction
were similar. The graph shows that for 50% correction
factor it takes 4 boards to reach the target registration. On Figure 10a. X and Y offset feedback correction with a 25%
the other hand, for 25% correction factor it takes about 8 correction factor applied to the feedback loop
boards to reach the target registration. As you would predict,
a smaller correction factor is slower to react but is less
sensitive to noise in the system thus holding the offset value
to a much tighter range.
Figure 10b. Ө offset feedback correction with a 25% Figure 11a. X and Y offset for a printing process without
correction factor applied to the feedback loop feedback showing consistent non-zero offset print-to-print

Comparison Test Procedure

The objective of this test was to compare the print process
performance of a printer without the presence of closed loop
control with a printer using position feedback control. An
additional objective of this test was to evaluate the stability
of the controller for long manufacturing periods (i.e., to
simulate a production condition) by determining the ability
of the controller to maintain the print process close to the
target registration value.

The test was performed for 100 boards using a

commercially available type 4, lead free paste. The board
used for this test was a commercial cell phone board, with
four phones to a panel. Due to the proprietary nature of the Figure 11b. Ө offset for a printing process without
product, the actual image of the board is not shown here. feedback showing consistent non-zero offset print-to-print
The board was 191mm X 117mm X I17 mm with OSP pad
finish. Four areas on the board were chosen to be monitored
for X, Y and Ө convergence. The four areas of the board
were chosen to provide adequate coverage of the board. As
a baseline at the beginning of the test, the board was aligned
to the stencil by the operator using visual methods. Once the
optimum alignment was achieved, the process was run for
100 boards without any positional feedback.

After the 100 board lot of boards was printed without

positional feedback, the feedback loop was turned on with a
correction factor of 50% .

Comparison Test Results

Figures 11a and 11b shows results from the baseline test
(no feedback employed) for F2R stroke direction only. R2F
stroke direction showed similar behavior. It is clear from the
plots, both X and Y offset fluctuate around a fixed offset
which is not zero.
Figure 12a. X and Y offset for the front-to-rear squeege
direction using feedback control showing consistent near
zero offset print-to-print

The results of turning on the feedback loop are shown in

figures 12 and 13. We see from these results the closed loop
control algorithm is capable of reaching the target value
rather quickly. Additionally, it is observed that the average
registration value can be maintained close to the target for
the entire duration of the test.

Figure 12b. Ө offset for the front-to-rear squeege direction
using feedback control showing consistent near zero
rotational offset print-to-print
Figure 13a. X and Y offset for the rear-to-front squeege
direction using feedback control showing consistent near
zero offset print-to-print
Figure 13b. Ө offset for the rear-to-front squeege direction
using feedback control showing consistent near zero
rotational offset print-to-print
Finally, comparing print performance between with/without
closed loop control shows process improvements produced
by the closed-loop control system. Comparison of the
process capability index, Cpk, for X and Y offset
measurements for the cell phone board is shown in Table 1.
The increase in the Cpk for both X and Y offset with control
is due to the centering and tightening of the print process.
Additional statistical analysis confirms the improvement of
print performance by employing the closed loop process
control. This analysis is presented in figure 14. Figure 14
shows the individual moving chart for X offset with both
with and without feedback control. It is clear from this
analysis that both mean and control limit improve for
process with the closed loop control.
Table 1. Cpk for the printing process alignment for without
and with positional feedback.
Cpk
Test Condition X offset Y offset
Without Feedback 1.6 1.5
With Feedback 2.0 2.3
Figure 14. Control Charts of a printing process without and
with feedback control
CONCLUSIONS AND RECOMMENDATIONS
Closed-loop controls have been implemented at many stages
along circuit board manufacturing lines. They have
successfully been implemented within reflow ovens and at
the components placement stage. As the push towards
product miniaturization becomes inevitable, it is clear that
closed loop-process controls for printing process will slowly
but surely required to maintain quality and yield of the
board assembly process. Closed-loop controllers, when
implemented correctly, present the advantages of keeping
complex processes within control limits even when small
external perturbations affect the product line. In addition,
closed loop control minimizes operator intervention and has
self-tuning properties.
From our limited laboratory environment, controlled
experiments, we have shown an improvement in the print
process capability using closed loop control. The full extent
of the benefit can only be accessed by employing such a
system in a true high volume production environment.